So consider the image below:
The CCD chips have AR on them which, going by the graph, work up to a point, that being the angle of incidence at which the AR no longer works.
The green lines in the image roughly show the cut off point.
I found that the photograph was taken at the Ball Aerospace facility and all 3 (or 4) photographs seem to match that facility with both staff and ceiling details.
The smaller colored lines indicate multiple refractions or reflections of the ceiling grid.
The orange lines indicate the angle of the secondary reflections inside the facility.
The red lines show the angles where the chips do not have AR and do not have the sapphire lens on top, so no curvature and no ceiling reflection at their angle, but do reflect the facility as the secondary reflection in the CCDs.
The curvature of both the grid reflected at least twice and of the secondary reflection in each block is due to the curvature of the sapphire lens that covers each CCD pair, as shown in the diagram.
Reason for only one side of each pair showing the features is that each pair is placed so that its sides are opposing each other.
Hope that's on the right track.